The European angle: thin in this evidence, and we say so
This piece is written for a European business audience, so we start with what the evidence does not contain. Our verified material has no EU regulation, no EU company data and no figures on European markets. We will not invent a regulatory story to fit the brief. What the evidence does show is a global pattern that European boards face too: a technology everyone intends to adopt, held back by the state of their own data.
The headline number, and a caution about its source
According to a report titled "Scaling AI agents with trustworthy data", 100% of respondents plan to use agentic AI within two years, and 69% expect to use it widely.1 The same survey puts AI access to company data at about 45% on average. In organisations it labels 'data laggards', that access falls to 30% or less.1 Only about half of respondents trust their agents' decisions, according to our summary of the Google Cloud survey behind the report.1
The source needs a warning. Our measured reliability check found that 0% of 11 checked claims from this MIT Technology Review document held up.1 That is a small sample, but it is a poor score, and a reader should treat these survey figures as indicative, not settled. The direction (high intent, weak data foundations) matches what the company announcements below describe. The exact percentages do not have independent confirmation here.
Where the money and partnerships are going: governance
Manulife and Microsoft announced a five-year agreement on July 22, 2026. Manulife will adopt Microsoft's Frontier Suite, deploy Microsoft Agent 365 and expand Microsoft 365 Copilot to more than 30,000 employees.2 The release frames this as AI governance as well as innovation. Shamus Weiland said: "Our partnership with Microsoft is a critical enabler of Manulife's continued evolution into a truly AI-driven organization."2
On July 21, 2026, Box announced new controls for AI agents working with enterprise content. They include agent guardrails, oversight of third-party agent activity, prompt injection detection and access policies based on agent classification.3 Tatsutoshi Murata of Nomura Research Institute said: "As we rapidly advance our utilization of AI agents, we expect Box—which has consistently led the development of security management capabilities for secure collaboration—to provide the administrative features needed to safely leverage this new era of AI."3
Both announcements come from company press releases carried by NewsEOD. Our measured reliability for that source is 57% of 4,956 checked claims holding up.2,3 Read them as what the companies say they are doing, not as results.
The picture from the chip maker
Nvidia sits behind much of this build-out. Its entity graph includes agent tooling such as NeMo Agent Toolkit and NeMo Guardrails, and customers such as Mount Sinai Health System and Yum! Brands.4 We can say what its filings show, and we cannot say what it earns from agents. The verified figures are:
Nvidia's cost of revenue was $16.621 billion in fiscal 2024, $32.639 billion in fiscal 2025 and $62.475 billion in fiscal 2026.5 Each year is roughly double the one before, our arithmetic on the filed figures. Cost of revenue is what it costs the company to make and deliver what it sells, so this tracks the scale of the hardware business. Cash on hand rose from $7.28 billion in fiscal 2024 to $8.589 billion in fiscal 2025 and $10.605 billion in fiscal 2026.6 These are SEC-filing figures. They show a company growing fast. They do not show that its customers are getting returns from agents, which is the question the survey raises.
The vertical-automation pitch: selling labour, not software
A recurring theme in CB Insights' interview series is that agent vendors size their market as labour spend. Covecta's Ben Thomas said: "Covecta's total addressable market (TAM) is tens of thousands of financial institutions globally, and for them we are not just disrupting their software budget but their labor budget as well."7 Covecta serves banks, non-bank lenders, building societies, credit unions and private credit firms, currently in the US and UK, with plans to go global.7 UK building societies and credit unions put part of this in a market European readers know.
Penguin AI's Glenn Herzberg makes the same move in healthcare, saying: "US Healthcare Administration runs about a trillion dollars a year, about a quarter of the total health spend, and the published estimates put around $570 billion of that in work that has no effect on health outcomes."8 That is a vendor's market-sizing claim, not an audited figure.
David Villalon of Maisa AI defines his market as "process automation of core business and production tasks at regulated industries", and says the work must be auditable, reproducible and hallucination resistant.9 For European buyers the 'regulated' and 'auditable' wording is the useful part. Our evidence does not say which regulations apply, so we draw no conclusion about EU rules.
Investors are mapping the category as well as funding it
CB Insights published a run of enterprise-AI interviews in late September 2026, with the CEOs of Shepherd, Arlo, LARX and Binary World and the CTO of Veridox.10 It also distributes datasets through Nasdaq Data Link. That tells us the category is being mapped and sold as a product, but it is not evidence that the companies interviewed are succeeding. On the funding side, Latitude, a payments platform, raised a $35M Series A on September 10, 2026.11 Investor Emily Man of Primary described backing Casap, a disputes company, because its founders had "both experienced the pain points of disputes firsthand at their respective large fintech companies".12
Via News's analysis
Our reading is that the evidence points to a gap between intent and readiness. The survey says everyone plans to deploy agents. Its own data access figures say the underlying data is largely out of reach, and the money in this dossier flows to governance (Manulife, Box) as well as to capability. The weakest link is measurement. Nothing here gives revenue or return on investment from deployed agents, and the one survey that quantifies readiness comes from a source that scored poorly on our checks.
What to watch
- Whether surveys of data access move above the 45% average, and from a source that holds up under checking.
- Whether the Manulife rollout to more than 30,000 employees produces reported results, not only announcements.
- Whether any European company or regulator figures arrive that let us write the EU story this brief calls for.
- Nvidia's cost of revenue and cash in the next filings, as a read on whether the build-out is still doubling.
